The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input via music generation prompts and local lyrics files.
Ingestion points: User-provided prompts in hanolab generate --prompt and content from local files via --lyrics-file (SKILL.md).
Boundary markers: The 'Boundaries' section explicitly instructs the agent to treat all names, filenames, and server responses as data, never as instructions (SKILL.md).
Capability inventory: The skill utilizes local terminal execution for the hanolab CLI and file system access (SKILL.md).
Sanitization: Recommends shell-quoting all paths and using structured process arguments to prevent injection (SKILL.md).
[COMMAND_EXECUTION]: Utilizes the hanolab CLI for audio generation and processing. The workflows are structured around specific command-line flags and avoid the use of dangerous shell features or arbitrary command construction.
[EXTERNAL_DOWNLOADS]: References the official HanoLab CLI guide and source code on GitHub. These references point to the vendor's own verified infrastructure and are intended for pre-installation or verification by the user.
[CREDENTIALS_UNSAFE]: Proactively addresses credential safety by instructing the agent never to request or display the HANOLAB_API_KEY. It directs the user to perform authentication via secure, local methods outside the agent's context.

hanolab